interplay between equilibrium and kinetics in prograde

Interplay between equilibrium and kinetics in progrademetamorphism of pelites: an example from the Nelsonaureole, British Columbia

D. R. M. PATTISON1 AND D. K. TINKHAM2

1Department of Geoscience, University of Calgary, Calgary, Alberta, T2N 1N4, Canada ([email protected])2Department of Earth Sciences, Laurentian University, Sudbury, Ontario, Canada

ABSTRACT The distribution of metapelitic mineral assemblages in the Nelson aureole, British Columbia, generallyconforms to what is predicted from phase equilibria. However, in detail, the sequence and spacing ofisograds, mineral textures and mineral compositions and mineral chemical zoning do not. Two of themain disequilibrium features in the aureole are: (i) delay in the onset and progress of several reactions,i.e. overstepping in temperature; and (ii) unreactivity of staurolite and especially garnet porphyroblastswhen they are reactants in prograde reactions. The thermal overstepping is ascribed to difficulty ofnucleation of the product porphyroblasts and sluggishness of dissolution of porphyroblasts when theyare reactants. The extent to which these kinetic barriers delay the onset of reaction is related to thereaction affinity of each reaction, defined herein as the Gibbs free-energy difference between thethermodynamically stable, but not-yet-crystallized, products and the metastable reactants. For overstepsin temperature (DT), reaction affinity is, in turn, related to the difference in entropy (DS) between thesetwo states through the relation A = DT * DS. Mineral reactions which release large quantities of H2O,such as chlorite-consuming reactions, have a higher entropy change per unit of temperature overstep,and therefore a higher reaction affinity, than those which release little or no H2O, such as the chlorite-free staurolite-consuming reaction. Thermal overstepping is consequently expected to be less for theformer than for the latter, as was estimated in the aureole where 0 to 30 �C overstepping was requiredfor garnet, staurolite and andalusite growth from a muscovite + chlorite-bearing precursor rock and�70 �C overstepping was required for the growth of Al2SiO5 from a staurolite-bearing, chlorite-freeprecursor. In all cases, reaction progress was strongly influenced by the presence or absence of fluid, withpresence of fluid lowering kinetic barriers to nucleation and growth and therefore the degree of thermaloverstepping. Textural features of rocks from the nearly coincident garnet, staurolite and andalusiteisograds are suggestive of a fluid-catalysed �cascade effect� in which reaction took place in a narrowtemperature interval; several competing muscovite + chlorite-consuming reactions, some metastable,appear to have occurred in parallel. Metamorphic reaction, fluid release and possibly fluid presence ingeneral in the aureole were episodic rather than continuous, and in several cases well removed fromequilibrium conditions. The extent to which these findings apply to regional metamorphism depends onseveral factors, a major one being enhanced deformation, which is expected to lower kinetic barriers tonucleation and growth.

Key words: equilibrium; kinetics; metamorphic reactions; overstepping; reaction affinity.

INTRODUCTION

This study addresses the interplay between theapproach to equilibrium and reaction kinetics duringprograde metamorphism of pelites. It expands on someof the ideas of Waters & Lovegrove (2002), whoargued for reaction overstepping on the order of30–40 �C for the production of staurolite and anda-lusite from a precursor assemblage of muscovite,chloritoid and chlorite in pelites from the Bushveldaureole. They ascribed the overstepping to a combi-nation of sluggish porphyroblast nucleation and slug-gish dissolution of precursor chloritoid, laying special

emphasis on the former. Their study of the Bushveldrocks showed that: (i) thermally activated devolatil-ization reactions can be significantly overstepped evenwhen the observed mineral assemblages conform clo-sely to predictions from equilibrium thermodynamics,and (ii) the actual reactions involved in the formationof a stable equilibrium assemblage can be considerablydifferent from those predicted by equilibriummodelling.

This study expands on some of these ideas using theNelson aureole, British Columbia (Pattison & Vogl,2005) as an example. We focus on two main issues: (i)contrasts in the degree of overstepping in temperature

J. metamorphic Geol., 2009, 27, 249–279 doi:10.1111/j.1525-1314.2009.00816.x

� 2009 Blackwell Publishing Ltd 24 9

of pelitic reactions as a function of reaction affinityand fluid presence; and (ii) the relative unreactivity ofporphyroblasts, in particular garnet and staurolite,when they are reactants in prograde reactions. �Reac-tion affinity� as used in this paper refers to the Gibbsfree energy potential that builds up when a reactionboundary is overstepped in pressure or temperature. Itrepresents the Gibbs free energy difference between thethermodynamically stable, but not yet crystallized,products and the metastable reactants. This energy isused to overcome kinetic barriers to nucleation andgrowth of the product assemblage. For oversteps intemperature (DT), reaction affinity is, in turn, relatedto the entropy difference (DS) between these twostates, through the relation A = DT * DS, whereDT = T – Teq. Reaction affinity, per unit of temper-ature overstep, is therefore higher for high-DSreactions, such as devolatilization reactions, than forlow-DS reactions, such as solid–solid reactions. For thesame energetic threshold for reaction, high-DS reac-tions are predicted to be overstepped less than low-DSreactions.

In this study, we contrast the formation of stauroliteand andalusite from a muscovite + chlorite-bearingprecursor assemblage (high DS, non-porphyroblasticreactants, low predicted thermal overstepping) with theformation of andalusite from a chlorite-free, staurolite-bearing precursor assemblage (low DS, porphyroblasticreactants, higher predicted overstepping). The signifi-cant overstepping of initial garnet formation from amuscovite + chlorite-bearing precursor assemblage isalso discussed. It is argued that, in all cases, fluid pres-ence, whether introduced or generated internally, had amajor effect on the degree of thermal overstepping. Atthe end of the paper, some of the broader implications ofour findings are discussed.

THE NELSON BATHOLITH AND AUREOLE

The Nelson Batholith is one of the largest of a suite ofmiddle Jurassic plutons in south-eastern BritishColumbia (e.g. Archibald et al., 1983). The batholith isemplaced in Triassic–Jurassic graphitic argillaceousrocks that were regionally deformed and metamor-phosed to biotite zone conditions in the Jurassic, priorto intrusion of the batholith.

The composite batholith ranges from tonalite togranite and comprises a northern mass, including theMt. Carlyle Stock, of about 30 · 50 km2 (referred toas the �main body�) and a 25 km long southern �tail�(Little, 1960; Vogl & Simony, 1992) (Fig. 1a). U-Pbdating indicates that the different phases of thebatholith were intruded in the Jurassic between c.173 and 159 Ma (Parrish, 1992; Sevigny & Parrish,1993; Ghosh, 1995). Geophysical and petrologicaldata indicate that the batholith is a tabular, flat-bottomed body, 2 to 7 km below the present-day sealevel (Cook et al., 1988). The batholith and aureoleare tilted �10� to the west, with the tilting ascribed

to a combination of Jurassic–Cretaceous contrac-tional deformation and Eocene extension (Pattison &Vogl, 2005).A 0.7–1.8 km wide contact aureole surrounds the

batholith. Different prograde metapelitic mineralassemblage sequences are developed in the argillaceoushost rocks around the aureole in a regular pattern,reflecting different pressures of contact metamorphismarising from the post-metamorphic tilting (Pattison &Vogl, 2005). Higher-pressure staurolite ± andalusite-bearing mineral assemblage sequences are restricted tothe aureole surrounding the eastern half of thebatholith, whereas lower-pressure cordierite ± anda-lusite-bearing mineral assemblage sequences arerestricted to the aureole surrounding the western halfof the batholith and its northern and southern tips(Fig. 1a). Despite the differences in mineral assem-blage sequences in the aureole, the total pressuredifference between them is only about 1.0 kbar(Fig. 1a; Pattison & Vogl, 2005). U-Pb dating ofmonazite that was produced during contact meta-morphism is the same age, within error, as the intru-sion (Tomkins & Pattison, 2007).

PROGRADE METAMORPHISM IN THESTUDY AREA

Figure 1b shows the part of the aureole examinedin this study, corresponding to Area D of Pattison& Vogl (2005), but showing more isograds andincorporating more recent sampling. A generaldescription of the metamorphic zones, accompaniedby photomicrographs, is provided in Pattison & Vogl(2005).The sequence of mineral-in isograds in the study

area is garnet, staurolite ± andalusite, sillimanite,K-feldspar, with the garnet, staurolite and andalusiteisograds coincident in some places at the level ofthe spatial resolution allowed by the sampling(Fig. 1b). Migmatization, interpreted to be due topartial melting, is associated with the development ofK-feldspar (Pattison & Vogl, 2005). The aboveprograde sequence is a classic facies series 2b (stauro-lite–andalusite) sequence in the scheme of Pattison &Tracy (1991), and represents a pressure of formation of�3.5–4.0 kbar assuming the Pattison (1992) Al2SiO5

triple point or �2.5–3.0 kbar assuming the Holdaway(1971) triple point.Below, the salient features of each zone are

described, with emphasis on the textures of garnet,staurolite and andalusite and the chemical zoning ingarnet. These observations are then compared withpredictions from thermodynamic modelling later in thepaper. Figure 2 shows garnet and staurolite textures atdifferent grades in the aureole. Figure 3 shows texturesof rocks from the andalusite and sillimanite zones.Figures 4 to 10 show garnet chemical zoning in rocksfrom the garnet, staurolite, andalusite, sillimanite andK-feldspar zones.

2 50 D . R . M. PA T T I S O N & D . K . T I N K H A M

� 2009 Blackwell Publishing Ltd

Mineral compositions and chemical zoning

Two or three representative garnets from each rockwere compositionally mapped and quantitativelyanalysed. These were typically the largest garnets inthe thin section. Garnet was mapped for Fe, Mn,Mg, Ca and Y on the JEOL JXA-8200 electronmicroprobe at the University of Calgary under thefollowing conditions: focused beam; 15 kV; 500 nA;100 ms dwell time; 2 lm pixel size. Quantitativegarnet analyses were performed using wavelength-dispersive analysis under the following conditions:focused beam; 15 kV; 50 nA; peak and backgrounddwell times of 20 and 10 s, respectively, for Fe, Mn,Mg, Ca, Al, Si and 40 and 20 s, respectively, for Y,P and Ti. Compositions of other minerals (biotite,muscovite, staurolite and plagioclase) are provided inPattison & Vogl (2005).

Regional metamorphic rocks outside the aureole

The host rocks outside the aureole in the study areaare regionally deformed and metamorphosed gra-phitic, locally sulphide-rich, argillites of the Triassicto Early Jurassic Ymir Group (Cairnes, 1934; Little,1960; Archibald et al., 1983; Pattison & Vogl, 2005).The argillites are variably interbedded with quartzite,siltstone and dark, graphitic marble. The regionalmetamorphic grade is the biotite zone. Most of theargillites outside the aureole contain biotite, butnone contains garnet. The rocks also contain chlo-rite, quartz, muscovite, plagioclase and graphite, inaddition to some or all of ilmenite, tourmaline,zircon, monazite, pyrrhotite, xenotime and allanite.Whole-rock chemical analysis of the argillites showsa narrow compositional range (Pattison & Vogl,2005).

Fig. 1. (a) Regional map of the Nelson batholith and aureole, adapted from Pattison & Vogl (2005). Dashed lines separatemineral assemblage domains of different pressure (see Pattison & Vogl, 2005; for details). (b) Mineral assemblages and isogradsfrom the study area. Sample numbers in boxes – samples analysed chemically and presented in Figs 4–10. Sample numbers withoutboxes – samples for which photomicrographs are presented in Figs 2 & 3.

E Q U I L I B R I U M AN D K I N E T I C S I N P R O G R AD E M E T AM O R PH I S M 25 1


Garnet zone

The garnet zone in Fig. 1b is narrow (<50 m) tonon-existent, within the resolution of the sampling,even though garnet is widespread in assemblagesat higher grade and from textural features formed

before the other porphyroblasts. The idealized reac-tion that forms garnet, in the model pelitic chemicalsystem KFMASH (K2O-FeO-MgO-Al2O3 SiO2-H2O),is:

Msþ ChlþQtz ¼ Grtþ Bt þH2O: ð1Þ

2.0 mm

91CW11

St

St

St

(e)

Gr t

1.0 mm 91CW9b (f)

Gr t

Gr t

St

St

Sil + Kfs zone

(d) 93CW19

0.5 mm Sil zone

(c) 93CW16

0.5 mm

Grt zone

(a) 93CW4

0.5 mm St zone

(b) 03Y C1

0.5 mm

Fig. 2. Textures of garnet (a–d) and of staurolite (e–f). Sample localities shown in Fig. 1b. See text for discussion of images.



Fig. 3. Textures of andalusite- and sillimanite-bearing rocks. Sample localities shown in Fig. 1b. See text for discussion of images.



Fig. 4. Photomicrographs, compositional maps and compositional analyses of garnet from the garnet zone, sample 03CW5 (seeFig. 1b for location of sample).



Figures 2, 4 and 5 show photomicrographs andgarnet chemical zoning maps of two rocks from thegarnet zone (03CW5 & 93CW4; Fig. 1b). Although93CW4 is physically within the staurolite zone, itcontains no staurolite or primary chlorite, suggestingthat the chlorite needed for staurolite formation wasconsumed beforehand during the formation of thegarnet in reaction (1).

Garnet is restricted to phyllosilicate-rich layers in03CW5 whereas it is more evenly distributed throughthe rock in 93CW4. The garnet in both samples showsmacroscopically rational crystal face development, butthat in 03CW5 is relatively inclusion-free whereas thatin 93CW4 has inclusion-filled rims. In thin section aconspicuous internal textural feature in garnet fromboth samples, especially in 93CW4, is a radial, hex-agonal growth pattern outlined by fine inclusions ofquartz, graphite and accessory phases (Figs 2 & 5b,c).Such a pattern has been observed in a number of otherlocalities (e.g. Andersen, 1984; Burton, 1986; Rice &Mitchell, 1991; Jamtveit & Andersen, 1992; Yardleyet al., 1996; Wilbur & Ague, 2006) and has beenascribed to porphyroblast growth under hydrostaticconditions in graphitic rocks, possibly accompanied byrelatively rapid growth. This distinctive pattern is dis-cernible in garnet at all grades (Fig. 2a–d), suggestingthat the mechanism of garnet growth even in high-grade rocks was similar to that now seen in rocks fromthe garnet zone.

The radial inclusion texture is accompanied by asubtle radial pattern in Ca concentration in which thin,outwardly emanating, flame-like zones of slightlyhigher and lower Ca concentration extend across themore dominant concentric zones of differing Ca con-centration (Figs 4 & 5). The origin of these radialtextures is unknown but, by analogy with the radialinclusion patterns noted above and with sector chem-ical zoning observed in igneous minerals, it may also bean indicator of relatively rapid growth of garnet.

In both samples, the dominant feature of Ca con-centration is an irregular, broadly oscillatory zoningfrom generally higher values in the centres to lowervalues at the rims (Figs 4 & 5). Explanations for thistype of oscillatory zoning are several (e.g. Chernoff &Carlson, 1997; Meth & Carlson, 2005), but the zoningsuggests processes more complex than those predictedfrom simple thermodynamic modelling, as discussedbelow. Yttrium zoning varies both between and withinthe samples, suggestive of local (sub-millimetre) vari-ations in Y concentration and lack of thin-section scaleequilibration.

In the garnet from 93CW4, Mn shows a typical bell-shaped pattern of decreasing Mn concentration awayfrom the core. Mg ⁄ (Mg+Fe) shows a slight increasefrom core to rim followed by a drop at the outer rim(Fig. 5g). The situation is similar in the garnet from03CW5 except that within a few tens of microns of therims there is a sharp increase in both Mn concentrationand Mg ⁄ (Mg+Fe).

The lack of other porphyroblasts in the rocks and theeuhedral texture of the garnet indicate that thesepatterns are due to growth rather than resorption. Apossible explanation for prograde Mn-enrichment atthe rims of the garnet in 03CW5 is consumption ofilmenite during garnet growth. Ilmenite with Mn ⁄(Mn+Fe) = 0.18 occurs as inclusions in garnet coresbut is absent from the matrix, whereas rutile does notoccur as garnet inclusions but is present in thematrix. Inthis pyrrhotite-bearing, S-rich rock (0.49 wt% S), thereaction of ilmenite to rutile may have been accompa-nied by the reaction of pyrite to pyrrhotite by the cou-pled Fe-, S- and Ti-conserving reaction, Ilm + Py =Rut + 2 Po (Connolly & Cesare, 1993). The Mg-enrichment in the garnet rims could also be explained bythis reaction, in which some of the Fe sequestered bypyrrhotite may have come from co-reacting silicateminerals like chlorite, resulting in concomitant growthof relatively magnesian garnet.

Staurolite zone

A separate staurolite zone, defined by rocks containingstaurolite but lacking andalusite, cannot be identifiedbecause the lowest grade occurrences of andalusite andstaurolite are in the same outcrops (Fig. 1b). However,there is a substantial, 800–1200 m wide domainupgrade of the staurolite ± andalusite isograd inwhich staurolite-bearing, andalusite-free mineralassemblages are common. Textural evidence indicatesthat staurolite growth preceded andalusite growth.Consequently, we refer to the abundant staurolite-bearing, andalusite-free rocks as belonging to thestaurolite zone for simplicity.

Figures 6 and 7 show photomicrographs and mapsof garnet chemical zoning of two representativeandalusite-free rocks from the staurolite zone(92CW10 & 91CW9c; Fig. 1b). The staurolite in92CW10 is unaltered whereas many grains of staurolitein 91CW9c are partially replaced by muscovite inter-grown with lesser amounts of biotite (Fig. 7a).

Garnet appears generally euhedral–subhedral in92CW10, although corners and edges of crystals in thematrix show evidence of rounding (Fig. 6a,b). Incontrast, garnet enclosed within staurolite shows sharpcorners (Fig. 6c). In 91CW9c, all garnet is sharplyeuhedral (Fig. 7).

Ca zoning in garnet from 92CW10 is dominated byconcentric oscillatory-type zoning, with a �0.05 mm-wide annulus of slightly higher Ca concentration at themargins. Superimposed on this pattern are the samepatchy, flame-like features seen in the garnet zone.Yttrium mimics the concentric Ca patterns, showinghigh core values and low marginal values, with anarrow annulus of elevated concentration within thelow-Y margin. The origin of this Y pattern wasascribed to successive breakdown of xenotime (high-Ycore) and allanite (Y annulus) during garnet growth(Tomkins & Pattison, 2007). In 91CW9c, Ca zoning



Fig. 5. Photomicrographs, compositional maps and compositional analyses of garnet from the garnet zone, sample 93CW4 (seeFig. 1b for location of sample).



Fig. 6. Photomicrographs, compositional maps and compositional analyses of garnet from the staurolite zone, sample 92CW10(see Fig. 1b for location of sample).



Fig. 7. Photomicrographs, compositional maps and compositional analyses of garnet from the staurolite zone, sample 91CW9c (seeFig. 1b for location of sample).



also shows concentric variations but none of the radialfeatures seen in 92CW10, and Y mimics Ca. Theidealized reaction forming staurolite is:

MsþGrtþ ChlþQtz ¼ Stþ BtþH2O: ð2ÞHowever, textural evidence for the participation ofgarnet in this reaction is minimal. In 92CW10 (Fig. 6),even the garnet showing rounded corners and edges,which might be taken as evidence for dissolution,shows internal growth zones defined by Ca and Y thathave similar rounded corners and edges, suggestingthat the crystals may never have developed sharpcorners and edges. The best evidence for dissolutioncomes from the truncation of Y and Ca zoning on theleft side of the crystal in Fig. 6d,e. Despite the evidencefor dissolution, there is no evidence for Mn-enrichmentat this edge. In 91CW9c, the textures and chemicalzoning suggest no garnet dissolution whatsoever. Wetherefore conclude that the reaction that best accountsfor staurolite development is:

Msþ ChlþQtz ¼ Stþ BtþH2O: ð3Þ

Andalusite zone

The location of the andalusite isograd is indistin-guishable from that of the staurolite isograd within theresolution of the sampling. The andalusite zoneextends 900 m up to the sillimanite zone. Andalusiteoccurs in three main textural varieties (Fig. 3).1 In samples from the lower grade part of the anda-lusite zone (e.g. 91CW8b in Fig. 1b), it grows asanhedral poikiloblasts scattered through the rock,accompanied by staurolite porphyroblasts of similaranhedral poikiloblastic texture and by small euhedralgarnet porphyroblasts (Fig. 3a).2 In the middle and upper andalusite zone, andalusitecommonly occurs as discrete, chiastolitic porphyro-blasts, sometimes with inclusions of euhedral staurolite(Figs 3b & 8a of this study, and fig. 6b of Pattison &Vogl, 2005). Staurolite porphyroblasts in the matrix ofthe rocks are commonly euhedral as well.3 In the middle and upper andalusite zone, andalusiteoccurs as partial or complete pseudomorphs ofstaurolite porphyroblasts (Figs 3d–f).More than one variety of andalusite may occur in thesame rock (e.g. Fig. 3e of this study and fig. 6b ofPattison & Vogl, 2005).

Andalusite-free St+Bt+Grt assemblages arecommon within the andalusite zone and may occur inthe same outcrop and even in the same thin section asandalusite-bearing assemblages. An example is shownin Fig. 3c–e. In one part of the thin section, stauroliteshows no evidence for reaction whatsoever (Fig. 3c),whereas millimetres away, staurolite porphyroblastsmay be completely converted to pseudomorphs ofandalusite (Fig. 3d). Intermediate stages, consisting ofpartial pseudomorphing of staurolite by andalusite,occur between the two extremes (Fig. 3e). Electron

microprobe analysis shows that there is no composi-tional difference between the reacting staurolite andthe unreacted staurolite, suggesting kinetic control.

Figure 8 shows maps of garnet chemical zoningfrom a representative andalusite-bearing rock exhibit-ing andalusite textural varieties 2 and 3 (93CW22;Fig. 1b). Two garnets were mapped: one in the matrixand a second included in andalusite. Examining thematrix garnet (Fig. 8d–g), the Ca zoning of the innerpart is similar to that of the staurolite zone sample92CW10. In contrast, however, is a distinct outermostrim of lower-Ca garnet. Boundaries between the threeCa domains are spatially matched by changes in Ycontent (Fig. 8e), with a narrow Y annulus developedat the interface between the outer low-Ca rim and theinner higher-Ca domains.

The garnet included in andalusite (Fig. 8h–k) showssimilar Ca and Y zoning except that the outer domainof low Ca is restricted to the side of the garnet facingaway from the centre of the enclosing andalusite. Thisasymmetry most likely arises from the fact that the sideof the garnet closest to the middle of the enclosingandalusite crystal side was armoured fairly early onduring andalusite growth, in contrast to the side facinginto the matrix which was open to new, low-Ca,growth on its outer margin.

The idealized reaction forming andalusite is:

Msþ StþQtz ¼ AndþGrtþ BtþH2O: ð4ÞEvidence for this reaction is seen locally in replacementof staurolite by andalusite, and the development of aconspicuous low-Ca rim on garnet which is absentfrom andalusite-free, staurolite-bearing rocks. How-ever, andalusite varieties 1 and 2 seem to have grownwithout any significant dissolution of staurolite, whichmay reflect growth from matrix muscovite and chloritein a reaction such as:

Msþ ChlþQtz ¼ Andþ BtþH2O: ð5Þ

Sillimanite zone

The sillimanite isograd occurs 400 m from the intrusivecontact and is marked by small amounts(<<1 modal%) of fibrolitic and fine-grained sillima-nite (see Pattison, 1992; p. 426–427, for a discussion ofthe distinction between the two). The sillimanite occursin the matrix and on the margins of andalusite andstaurolite porphyroblasts (Fig. 3g).

Three hundred metres from the contact, the rocktexture changes abruptly, taking on a more schistose,less porphyroblastic appearance than at lower grade,and sillimanite is abundantly developed. Staurolite andandalusite porphyroblasts are replaced by coarsemuscovite commonly containing sillimanite needles(Figs 3h & 9 of this study and fig. 6e of Pattison &Vogl, 2005). The coincidence of textural change,staurolite consumption and abundant sillimanitedevelopment is shown in Fig. 1b as the �major stau-rolite-out, sillimanite-in� isograd. Staurolite relics



Fig. 8. Photomicrographs, compositional maps and compositional analyses of garnet from the andalusite zone, sample 93CW22(see Fig. 1b for location of sample).



Fig. 9. Photomicrographs, compositional maps and compositional analyses of garnet from the sillimanite zone, sample 93CW16(see Fig. 1b for location of sample).



Fig. 10. Photomicrographs, compositional maps and compositional analyses of garnet from the K-feldspar zone, sample 93CW19(see Fig. 1b for location of sample).



persist a little upgrade in some rocks (Fig. 1b). The lastobserved occurrence of andalusite is �150 m from thecontact, defining a zone of co-existence of And+Sil ofabout 250 m width.

Garnet in some rocks from the sillimanite zoneshows inner portions with radial inclusion patternssimilar to those seen in garnet from the garnetand staurolite zones (Fig. 2c). The rims commonlycontain needles of fine-grained sillimanite (Fig. 9b,c),indicating garnet growth during or after sillimanitegrowth.

Figure 9 shows maps of garnet chemical zoningfrom a representative sillimanite-bearing rock(93CW16; Fig. 1b). Two garnets were mapped: one inthe matrix and a second included in a muscovitepseudomorph (most likely after staurolite). Examiningthe matrix garnet (Fig. 9d–g), Ca zoning is similar tothat in andalusite zone sample 93CW22, including anoutermost rim of low-Ca garnet within which thesillimanite is included. Whereas in andalusite zonesample 93CW22 the interface between the high-Cadomain and the low-Ca outer rim is sharp and euhe-dral, in sillimanite-bearing sample 93CW16 the inter-face is irregular and embayed in places, suggestive oflocalized dissolution prior to growth of the outer rim.

Yttrium content matches the inner Ca zones(Fig. 9d,e), with an abrupt drop-off in Y matching thechange from moderate Ca values in the inner core tohigher values in the outer core. However, there is littlechange in Y concentration across the anhedral inter-face between the low-Ca rim and the inner higher-Cadomain where dissolution is inferred. The garnetincluded in the muscovite pseudomorph (Fig. 9h–k)shows broadly similar chemical zoning.

The development of abundant sillimanite in the rockmatrix and within the low-Ca garnet rims, concomitantwith major consumption of staurolite, suggests thefollowing reaction:

Msþ StþQtz ¼ SilþGrtþ BtþH2O: ð6ÞThe reaction responsible for the initial, modally minordevelopment of sillimanite in rocks downgrade of themain development of sillimanite is uncertain, but mayhave involved reaction of matrix minerals. Andalusiteshows no evidence of instability until the major stau-rolite-out, sillimanite-in isograd is crossed, at whichpoint it is variably replaced by muscovite and sillima-nite. Although there is little direct evidence for pro-gress of the polymorphic inversion reaction,

And ¼ Sil; ð7Þthis reaction may have been involved in the wholesaletextural and mineralogical changes occurring at themajor staurolite-out, sillimanite-in isograd.

K-feldspar zone

Within 100 m of the contact, K-feldspar occurs in theassemblage Ms+Qtz+Kfs+Sil+Bt±Grt, marking

the K-feldspar-in isograd. K-feldspar occurs inmedium- to coarse-grained leucosomes consisting ofanhedral, variably perthitic crystals intergrown withquartz and sillimanite (fig. 6g of Pattison & Vogl,2005). These features are suggestive of partial meltingby the reaction:

MsþQtz ¼ SilþKfsþmelt: ð8ÞGarnet in a rock from the K-feldspar zone shows

inner portions with radial inclusion patterns similarto those in garnet at lower grade (Fig. 2d), mantledby anhedral–euhedral rims containing needles of fine-grained sillimanite (Fig. 10b,c). Figure 10 showsmaps of garnet chemical zoning from this rock(93CW19; Fig. 1b). The zoning is more complex thanat lower grade, with Ca and Y defining the followingdistinct zones with abrupt boundaries: an anhedralinner zone, containing the highest Ca and Y con-centrations, corresponding to the zone of radialinclusions; a zone of lower Ca and Y, with a high-Yannulus at its outer edge; and an outer rim in whichCa concentration is slightly higher than the previouszone, but Y concentration is variable (compareFig. 9e,i). High Mn and low Mg ⁄ (Mg+Fe) on theanhedral outer edge of the garnet are consistent withlocal resorption.

PHASE EQUILIBRIUM MODELLING

Figure 11 shows phase diagram sections calculatedfor the average composition of pelites from theNelson aureole from Pattison & Vogl (2005), givenin Table 1. In terms of the Thompson (1957) AFMprojection, the bulk composition has an A-value of0.1 and an Mg ⁄ (Mg+Fe) of 0.46. The averagecomposition of eight pelites just from the study areaappears to be slightly Fe-richer [average Mg ⁄(Mg+Fe) = 0.44] and Mn-richer (0.092 wt% MnO v0.082 wt% MnO in the average composition), butwhether this difference is statistically significant isuncertain.

The thermodynamic data set used to calculatethe phase diagrams is that of Holland & Powell(1998) updated to ds5.5 in 2003. Activity modelsused in all calculations are those used in Tinkham& Ghent (2005) with the following exceptions: (i)margarite was not considered as a component inwhite mica, and (ii) the ternary feldspar model ofHolland and Powell (2003), using a molecularmixing model and asymmetric formalism, was usedinstead of separate plagioclase and pure orthoclaseor sanidine.

Two phase diagram software packages were used tocalculate the phase diagrams: THERMOCALCTHERMOCALC v. 3.25(Powell, 1998) and THERIAK-HERIAK-DOMINOOMINO (De Capitani &Brown, 1987; De Capitani, 1994). For the samethermochemical database and activity models,THERIAKHERIAK-DOMINOOMINO gives calculated phase boundaries towithin 1–2 �C of the THERMOCALCTHERMOCALC results.



Choice of chemical system for modelling

Four chemical systems were used to model the phaseequilibria (Fig. 11): MnNCKFMASHT (MnO-Na2O-CaO-K2O-FeO-MgO-Al2O3-SiO2-H2O-TiO2, with Cand P2O5 omitted from the whole-rock analysis,followed by projection from pyrrhotite); MnNCKF-MASH (system reduced from MnNCKFMASHTby projection from ilmenite); MnKFMASH (system

reduced from MnNCKFMASH by projection fromalbite and anorthite); KFMASH (system reduced fromMnKFMASH by omission of MnO). For each system,the molar proportions were normalized to 100%. Theanhydrous whole-rock compositions for each systemare listed in Table 1. The H2O component was addedto each chemical system in the phase equilibriummodeling, in the form of excess H2O. Fe3+ was notconsidered as a separate component in the modelling

Fig. 11. Phase diagram sections for the average Nelson pelite composition in chemical systems of increasing complexity, calculatedusing THERMOCALCTHERMOCALC (Powell, 1998) and a modified version of the 2003 update of the Holland & Powell (1998) data base (see text fordetails). Abbreviations from Kretz (1983). The bulk compositions for each diagram are presented in Table 1. Increasingly darkershades of grey represent domains of increasing Gibbs phase rule variancy, starting from white = divariant. The P–T box in (d) isexpanded in Fig. 12a.



because Fe3+ and Fe2+ were not measured separatelyeither in the whole-rock analyses or in the mineralanalyses. Most rocks in the aureole contain graphite,ilmenite and pyrrhotite, suggestive of not stronglyvarying, and probably low, amounts of Fe3+ (Con-nolly & Cesare, 1993). Although at pressures above theAl2SiO5 triple point the phase equilibria vary signifi-cantly between the different systems, in the pressurerange of interest to this study (2–5 kbar) the maindifference relates to the stability of garnet, and to alesser degree the stability of cordierite v. andalusite.

Comparison of the garnet-in line in the four differ-ent chemical systems in Fig. 11 shows that its positionis sensitive to bulk composition and possibly thermo-dynamic data, in contrast to the staurolite-in, andalu-site-in and staurolite-out lines whose positions do notchange very much from one chemical system toanother. We chose the MnNCKFMASHT system formodelling the phase equilibria because it allowsassessment of Mn and Ca variation in garnet, andtakes account of the effect of ilmenite in sequesteringsome of the Fe and Mn in the rocks. Compared withthe Ti-free MnNCKFMASH system, it predicts asmaller stability field of garnet, which agrees with thegeneral absence of garnet in low-pressure cordierite-bearing rocks in the Nelson aureole and in aureolesworldwide (Pattison & Tracy, 1991). On the otherhand, it is possible that the contracted garnet stabilityfield predicted in MnNCKFMASHT, in which garnet-in occurs only 10–15 �C below staurolite-in at pres-sures below the triple point, is too small when oneconsiders the lower-temperature positions of the pre-dicted garnet-in line for individual bulk compositionsfrom the study area (see below), and the fairly wide-

spread development of a distinct garnet zone down-grade of the staurolite isograd in many regional faciesseries 2b (staurolite-andalusite-type) sequences of thesame pressure as in this study (see compilation inPattison & Tracy, 1991).

Pressure of contact metamorphism

Figure 12a is an enlargement of the part ofMnNCKFMASHT phase diagram in Fig. 11dbetween 2 and 5 kbar and 500 and 650 �C. Thesequence of mineral-in reaction isograds in Fig. 1b canbe matched to the mineral-in reactions in Fig. 12a inthe pressure range 3.3–4.0 kbar, over which rangeSt+And is stable. We have chosen 3.5 kbar forsubsequent modelling because of the coincidence ofthe staurolite and andalusite isograds in the fieldand the significant separation between these and thesillimanite isograd, recognizing that these could beinfluenced by the kinetic factors discussed below. Theexact pressure does not affect the central arguments ofthe paper.

Temperatures of isograds

Table 2 lists the predicted temperatures of the reactionisograds along the 3.5 kbar isobaric line in Fig. 12a.The temperature ranges in Table 2 represent the min-imum and maximum temperatures for each isogradreaction between 3.3 and 4.0 kbar in MnNCKF-MASHT (Fig. 12a).

Figure 13a shows that even for the same chemicalsystem, the position of the garnet-in line varies signi-ficantly for different individual rock bulk composi-tions. In the light of this sensitivity, we have estimatedthe temperature of the garnet-in reaction to be 527 �Cwithin a range of 498–546 �C, based on the averageand range of the garnet-in reaction for the differentbulk compositions shown in Fig. 13a. The generallylower temperatures of the garnet-in line for the indi-vidual compositions plotted in Fig. 13a, comparedwith that for the average composition, may be due totheir slightly Fe+Mn-richer bulk compositions.

In Fig. 14 the isograds are plotted against distancefrom the contact in the southern part of the aureolewhere there is the densest sampling (i.e. outwards from93-CW-19 in Fig. 1b). Individual boxes in Fig. 14correspond to a distance uncertainty of ±25 m and atemperature uncertainty corresponding to the temper-ature range in Table 2, with the solid dots in the boxesbeing the temperature estimate for 3.5 kbar. Absolutetemperature uncertainties, arising from uncertainties inthe thermodynamic data, activity–composition rela-tions of the minerals, and possible deficiencies in thechoice of chemical system for the modelling (e.g. noconsideration of Fe3+), are likely larger than thetemperature ranges in Table 2, but are difficult toquantify (see Waters & Lovegrove, 2002 and Powell &Holland, 2008 for detailed discussions). However,

Table 1. Average Nelson whole-rock compositions used forphase diagram modelling.

Raw

analysisa

(wt%)

Raw

analysisb

(mol.%)

MnNCKFM

ASHTc

(mol.%)

MnNCKF

MASHd

(mol.%)

MnKFMASHe

(mol.%)

KFMASHf

(mol.%)

SiO2 60.41 68.53 69.89 71.05 71.76 71.84

TiO2 0.93 0.80 0.81

Al2O3 20.10 13.44 13.70 13.93 13.44 13.45

FeO 5.68 5.39 5.44 4.71 5.83 5.84

MnO 0.08 0.08 0.08 0.08 0.10

MgO 2.30 3.89 3.96 4.03 4.99 4.99

CaO 1.06 1.29 1.32 1.34

Na2O 1.53 1.68 1.71 1.74

K2O 4.17 3.02 3.08 3.13 3.87 3.88

P2O5 0.15 0.07

LOI 3.09

C 0.31a 1.77

S 0.027a 0.06

Total 99.51 100.00 100.00 100.00 100.00 100.00

aRaw analysis from Pattison & Vogl (2005): C and S are part of LOI so were not added to

total.bRaw molar proportions, with LOI omitted but C and S included, normalized to 100%.cMnNCKFMASHT: omit C and P2O5 from raw analysis, project from pyrrhotite,

renormalize to 100%.dMnNCKFMASH: as above, then project from ilmenite, renormalize to 100%.eMnKFMASH: as above, then project from albite and anorthite, renormalize to 100%.fKFMASH: as above, then drop MnO, renormalize to 100%.



relative uncertainties are unlikely to be much greaterthan those listed in Table 2 when one considers theconsistency of facies series 2b isograd sequencesworldwide and the small predicted temperatureseparation (10–40 �C) between the isograd reactions(Pattison & Tracy, 1991).

THERMAL MODELLING

Figure 14 shows a thermal profile for the aureole(maximum temperature v. distance from the contact),calculated using the one-dimensional thermal model-ling software of Peacock (1990). The tabular shape ofthe tail of the Nelson batholith adjacent to the studyarea (Fig. 1a) makes this simple modelling appropri-ate. The parameters used in the construction of the

Fig. 12. (a) P–T phase diagram section for average Nelsonpelite in the system MnNCKFMASHT (see Fig. 11d and text).Labels in the ellipses are for mineral-in and mineral-out reac-tions interpreted to correspond to the isograd reactions inFig. 1b. Intervals of predicted garnet growth and consumptionalong the inferred isobaric 3.5 kbar isobaric P–T path areindicated by thick and thinner solid lines, respectively. (b, c)Changes in predicted mole fractions of phases, normalized to asum of one molar oxide in each phase (� mineral modes), goingupgrade along the isobaric 3.5 kbar P–T path in (a).

Table 2. Temperatures and distances of isograds from thecontact.

Distance from

contacta (±25 m)

Temperatureb

(�C)Temp. rangec

(�C)

Grt-in 1400 527d 498–546d

St-in 1340 557 553–563

And-in 1340 560 553–578

Sil-in 390 611 578–630

Major St-out,

Sil-in

290 563 556–583

St-relics out 200 563 556–583

And-out 130 611 578–630

Kfs-in 70 653 646–668

aMeasured outwards from 93-CW-19 in Fig. 1b.bTemperatures of isograds at 3.5 kbar for MnNCKFMASHT.cTemperature range of isograds between 3.3 and 4.0 kbar for MnNCKFMASHT.dAverage and range of Grt-in line for different bulk compositions and chemical systems

for 3.5 kbar (see text for discussion).



profile are given in the caption to Fig. 14, based on apriori estimates of thermal and spatial parameterspertinent to the Nelson intrusion and country rocks.

The calculated thermal profile in Fig. 14 is never-theless non-unique and can vary substantiallywhen one or more of the several poorly constrainedinput parameters are varied, including some notincorporated in Peacock�s modelling such as the effectof endothermic metamorphic reactions. On the other

hand, all calculated profiles have the same form as theone in Fig. 14, so that comparison of the observedlocations of the isograds with this profile revealsimportant first-order features that are independent ofthe details of the thermal modelling. The profile shownin Fig. 14 is consistent with minimal overstepping ofthe muscovite-out ⁄K-feldspar-in and chlorite-out ⁄andalusite-in reaction isograds, a prediction that isdiscussed further below.

Fig. 13. (a) P–T equilibrium phase diagram section (same as in Fig. 12a) showing positions of garnet-in lines for bulk compositionsof individual rocks, and garnet–core–isopleth intersections for individual rocks whose garnet have values of XMn

Grt that occurwithin the range in (c) (see text for discussion). (b) Calculated isopleths of XCa

Grt, and range of measured XCaGrt in garnet cores and

rims from rocks in the andalusite and sillimanite zones. (c) Calculated isopleths of XMnGrt . (d) Calculated isopleths of

Mg ⁄ (Mg+Fe)Grt.



COMPARISON OF PREDICTED PHASEEQUILIBRIA WITH OBSERVATIONS

The following section compares the observed featuresof the prograde metamorphic sequence in the Nelsonaureole with the thermodynamically predicted features.This approach involves several assumptions. First,chemical equilibrium is assumed. Studies such asCarlson (2002) have provided evidence for thinsection-scale chemical disequilibrium during garnetgrowth, especially involving Ca and some traceelements, which they ascribed to the kinetics of inter-granular diffusion. For this reason, we only focus onmajor features of composition and chemical zoningthat can be plausibly related to reactions involving therock-forming minerals. Preserved growth zoning ingarnet and general lack of garnet dissolution indicatesthat prograde fractional crystallization has occurred.Whereas modelling of prograde fractional crystalliza-tion of garnet, discussed later in the paper, results insignificant contraction of garnet stability comparedwith the equilibrium model, it has relatively little effecton other phase boundaries or on mineral compositionisopleths, most likely because of the minor amounts ofgarnet (<1 vol.%) that are predicted to be formed.

Second, it is assumed that the reaction isogradtemperatures are close to those calculated in Figs 12and 13 and Table 2, acknowledging that there isuncertainty in the absolute and relative temperaturesdue to the factors discussed above, and to the assumed

pressure of contact metamorphism. For example,assuming a pressure of 3.8 kbar rather than 3.5 kbarresults in a larger predicted temperature separationbetween the staurolite-in and andalusite-in reactions(15 �C v. 3 �C), and a smaller predicted temperatureseparation between the andalusite-in and sillimanite-inreactions (20 �C v. 50 �C). The greatest uncertainty ison the temperature of the garnet-in reaction. Althoughit could be argued that, within the uncertainties of thecalculations, there is no significant temperaturedifference between the garnet-in and staurolite-inreactions, the common textural evidence for earlygarnet growth in the Nelson rocks, and the abundanceof prograde sequences of the same pressure as theNelson aureole that do contain a significant garnetzone downgrade of the staurolite zone (Pattison &Tracy, 1991), suggest that there is a significant tem-perature difference between them. In the interpreta-tions that follow, the consequences of varying thetemperature of the garnet-in reaction are discussed.Third, we assume that aH2O was unity during

progress of the prograde reactions, even though therocks are graphitic. Assuming the graphitic metapel-ites were not pervasively infiltrated with fluid duringintervals of reaction, when they were internally gen-erating and expelling their own H2O, aH2O will havebeen internally buffered by graphite to values of�0.85–0.95 (Connolly & Cesare, 1993). The result isan overall lowering by 10–15 �C of the absolutepositions of the dehydration reactions in Table 2, but

Fig. 14. Comparison of modelled thermalprofile in the Nelson aureole (solid line) withthe observed locations of isograds and theirtemperatures assuming equilibrium (boxeswith dots). For the thermal modelling, thesoftware of Peacock (1990) was used, with thefollowing parameters: ½ width of tabularintrrusion: 1750 m. T of country rock at timeof intrusion: 400 �C. T of magma: 900 �C.Tof crystallization interval: 200 �C.Enthalpyof crystallization: 100 kJ kg)1. Thermalconductivity: 2.25 W m K)1. Heat capacityof solids: 1000 J kg K)1. Density of solids:2.75 kg m)3. Time step: 500 a. Number oftime steps: 2000. Total time elapsed insimulation: 1 Myr. For the isograd locations,the dots and boxes represent the mean andrange of the equilibrium isograd temperaturescombined with a distance uncertainty of±25 m (data in Table 2).



an insignificant change in the temperature differencesamong them (Pattison, 2006).

Mineral-in isograds

The predicted order of mineral growth in Fig. 12aagrees with the porphyroblast textures (Figs 2–10), inwhich garnet occurs as inclusions in all porphyroblastsand staurolite occurs as inclusions in andalusite, withthe opposite relations never observed. A caveat to thisinterpretation is that such enclosing relationships showthe relative timing of the end of growth of porphyro-blasts rather than the beginning of their growth (e.g.Vernon & Powell, 1976), such that it is possible thatlarger porphyroblasts such as staurolite or andalusitecould enclose smaller, possibly later-nucleating, phaseslike garnet. We nevertheless favour the first interpre-tation because of the consistency of these enclosingrelationships in the Nelson aureole and in other faciesseries 2b (staurolite-andalusite) settings worldwide(Pattison & Tracy, 1991).

On the other hand, the predicted spacing of themineral-in isograds does not agree with the observedspacing (compare Figs 1b, 12a & 14). In the field, thegarnet, staurolite and andalusite isograds are almost

coincident (Fig. 1b), even though the garnet-in reac-tion is predicted to occur �30 �C below the staurolite-and andalusite-in reactions (Table 2). A drop of 30 �Cfrom conditions of the staurolite-in isograd equates toabout 900 m in distance (Fig. 14). When the temper-ature uncertainty on the garnet-in isograd is taken intoaccount, this distance ranges between �250 m and�2000 m. Figure 15 compares the mapped position ofthe isograds from Fig. 1b with the predicted positionof the isograds based on the thermal profile in Fig. 14and assuming that the reactions proceeded at equilib-rium along the 3.5 kbar isobaric line in Fig. 12a. Fig-ure 15 shows that there is potentially a substantialdomain in the field downgrade of the observed garnetisograd where garnet should have grown according toequilibrium predictions.

The spatial coincidence of the staurolite-in andandalusite-in isograds in Fig. 1b is consistent with thesmall predicted separation between the isograd reac-tions of <5 �C (Fig. 12a and Table 2), acknowledgingthe possibility that the two reactions could have beenmore separated in temperature but proceeded togetherbecause of kinetic factors. The sillimanite-in isogradoccurs �20 �C and 250 m upgrade of where it is pre-dicted to occur in Fig. 14. As noted above, the initial

Fig. 15. Comparison of observed isograd locations with predicted locations assuming equilibrium. (a) Observed isograds from Fig. 1b.(b) Predicted positions, based on temperatures in Table 2 and distances corresponding to those temperatures from the thermal profilein Fig. 14.



development of sillimanite involves small amounts offine-grained and fibrolitic sillimanite, and appears tobe due to localized reaction involving matrix mineralsrather than whole-rock reactions. The major sillima-nite-forming reaction, associated with major con-sumption of staurolite, is discussed below.

Mineral-out isograds

The predicted sequence and spacing of mineral-outisograds do not match the observations at all. Garnetis ubiquitous at all grades and shows little evidence forconsumption, even though it is predicted to dissolvesignificantly during staurolite production and to becompletely consumed downgrade of the Kfs+Sil iso-grad by the continuous reaction:

MsþGrt ¼ Andþ BtþQtz: ð9ÞThemajor staurolite-out, sillimanite-in isograd occurs�1050 m upgrade of where staurolite is predicted tohave reacted out, corresponding to a temperatureoverstep of �70 �C. Staurolite relics persist at evenhigher grade. In contrast, there is evidence for patchyreaction of staurolite to andalusite throughout the1100 m interval between the staurolite + andalusiteisograd and the major staurolite-out, sillimanite-inisograd (Figs 1b & 3).

The andalusite-out isograd occurs 500 m above thepredicted andalusite-out reaction, corresponding to a

temperature overstep of �45 �C. Whereas the meta-stable persistence of andalusite into the sillimanite zoneis well known in other contact aureoles (see Pattison &Tracy, 1991 for a review), the unreactive behaviour ofgarnet and staurolite is less commonly noted.

Textures and chemical zoning of garnet

The isobaric P–T path through Fig. 12a predictsintervals of garnet growth and dissolution for theaverage Nelson pelite composition that can be com-pared with the observed textures and chemical zoningof garnet in the aureole. In addition to the generalcaveats noted above, this approach fails in detailbecause small differences in bulk composition betweenindividual rocks and the modelled average composi-tion in Fig. 12a, including in minor elements such asCa, Mn and P, can have a significant effect on thestability and composition of the garnet. Nevertheless,in all models, garnet is predicted to undergo twointervals of growth and two intervals of consumptionduring prograde metamorphism (Fig. 12a), the lastleading to complete consumption in the andalusitezone by reaction (9) (for some individual bulk com-positions, final garnet consumption occurred in thesillimanite field). Predicted changes in mole fractions ofphases along the 3.5 kbar isobaric path, normalized toa sum of one molar oxide in each phase (and thereforea good approximation to modal changes), are provided

Fig. 16. Area v. volume in porphyroblastsize reduction, assuming an equidimensionalporphyroblast such as garnet. The largenumbers above each frame are per centvolume reduction going from the outerdashed outline to the inner solid outline,whereas the small numbers beneath eachframe are per cent volume increase goingfrom the inner solid outline to the outerdashed outline.



in Fig. 12b,c. Calculated isopleths of XCaGrt, XMn

Grt

and Mg ⁄ (Mg+Fe)Grt for the average pelite composi-tion are shown in Fig. 13b–d, and isoplethintersections for garnet cores from three sampleswhose XMn

Grt fall within the range in Fig. 13c, areshown in Fig. 13a.

Comparison of predicted modal changes with what isobserved in the rocks requires a quantitative estimate ofthe extent of garnet dissolution or growth frompetrography. This is difficult because a thin section is atwo-dimensional slice through a three-dimensionalobject, so that significant volume reductions may resultin only small reductions in area. The problem isillustrated in Fig. 16, in which volume reductions of upto about 20% may be difficult to detect. Figure 16 alsoshows that cut effects must be considered. In contrast tothe euhedral resorbed products illustrated in Fig. 16,dissolution is unlikely to proceed evenly on all sides of aporphyroblast, instead being focused at corners andedges of crystals, and in domains of enhanced reactiv-ity. Figure 17 shows a comparison of zoning in XCa

Grt

from the garnet, staurolite, andalusite and sillimanitezones with predicted garnet growth and consumptionfrom phase equilibrium modelling, in which modalchanges are assessed with reference to Fig. 16.

Garnet zone

Garnet growth is predicted over a �30 �C intervaldowngrade of the first development of staurolite.Extending the garnet isopleths in Fig. 13b–d to theinferred temperature of 527 �C at 3.5 kbar for theonset of garnet growth (�25 �C below the garnet-inline shown in Figs 12 & 13), XCa

Grt is predicted tosmoothly decrease from 0.09 to 0.06 through thisgarnet growth interval, ignoring minor fractionaleffects. Euhedral chemical zoning in garnet from thegarnet zone (Figs 4 & 5), and in the inner portions ofgarnet from the staurolite, andalusite and sillimanitezones (Figs 6–9), is consistent with progressive growthof garnet, but in contrast to the steadily decreasingvalues of XCa

Grt predicted in Fig. 13b, the oscillatory

Fig. 17. Comparison of texture and calcium zoning of garnet (top row), with predicted garnet growth and consumption for twoend-member models: equilibrium (second row), and perfect fractional growth of garnet with no dissolution (bottom row). Simplifiedmineral reactions for the equilibrium model, based on Fig. 12, are listed beneath the second row. See text for discussion.



and flame-like Ca zoning indicates more complexprocesses. The concentric Ca zoning could indicateinvolvement of minor Ca-bearing phases like epidoteand allanite that are not predicted to be present inFig. 12a, but which are observed in the rocks. Thispossibility is further suggested by the Y zoning, whichTomkins & Pattison (2007) argued reflected successiveconsumption of xenotime and allanite during garnetgrowth, the latter producing monazite. The range ofobserved values of XCa

Grt in the garnet centres (0.05 to0.07, neglecting the anomalously Ca+Mn-rich garnetin 03CW05) is lower than predicted by equilibriummodelling. This discrepancy could be ascribed to anumber of factors, such as errors in the thermody-namic data or small differences between the modelbulk composition and the bulk composition of theindividual rocks containing the analysed garnet. Onthe other hand, the clustering of garnet core isoplethintersections around the staurolite-in reaction inFig. 13a is consistent with the close coincidence of thegarnet, staurolite and andalusite isograds in the field,and could be accounted for by delayed garnet growth.

Staurolite zone

Approximately 70% of the garnet that grew in thegarnet zone is predicted to be consumed in theproduction of staurolite (Fig. 12c). The euhedralexternal shape and internal chemical zoning of garnet inthe staurolite zone (Figs 6&7) argues against significantresorption, although there is some evidence for round-ing of corners of some garnet crystals. Comparison withFig. 16 suggests that it is unlikely that there was morethan about 10% garnet resorption during staurolitegrowth, and possibly none at all in some rocks.

Andalusite zone

Figure 12c predicts a second interval of garnet growthaccompanying andalusite growth and stauroliteconsumption, such that it attains 89% of its maximumpredicted mode immediately prior to staurolitegrowth. Evidence in support of a second interval ofgarnet growth is the low-Ca rim that occurs on themargin of garnet grains in andalusite-bearing rocks(Figs 8 & 17). The width of the low-Ca outer rims inFig. 8, assuming near-centre cuts through the garnetand with reference to Fig. 16, suggests 40–70% in-crease in garnet volume accompanying development ofandalusite, much less than the �200% growth pre-dicted in Figs 12c & 17. The XCa

Grt of the rims is 0.02–0.03 lower than the garnet on which the rims grew, adifference that is significantly greater than the verysmall differences predicted in Fig. 13b. Assuming thatandalusite growth occurred primarily due to staurolitebreakdown, the lower than predicted XCa

Grt is con-sistent with the kinetically delayed staurolite break-down to andalusite noted earlier, in which the garnetthat grew as part of this reaction inherited a compo-

sition reflecting the higher temperatures of the kineti-cally delayed reaction.

Sillimanite zone

Garnet is predicted to be consumed in the andalusitezone (Fig. 12). However, garnet in sillimanite zonerocks (e.g. Fig. 9) has low-Ca rims that containsillimanite needles, indicating garnet growth in thesillimanite field. As above, the most likely explanationis metastable persistence of staurolite into the stabilityfield of sillimanite, at which point it reacted to formsillimanite and low-Ca garnet (i.e. the staurolite-out,major sillimanite-in isograd in Figs 1b & 14). Thelow-Ca rims of garnet in Fig. 9 show a somewhatirregular margin on their inner edge, suggestingdissolution prior to growth of the rim, possibly inassociation with staurolite growth by reaction (2), orthrough reaction (9).

K-feldspar zone

Garnet in this zone contains a relatively high-Ca coremantled by a low-Ca rim, like that seen in the anda-lusite and sillimanite zones, but contains in addition anouter low-Ca rim, separated from the inner low-Ca rimby a euhedral Y annulus (Fig. 10). Sillimanite needlesoccur in the outer rim. Assuming that the core andinner low-Ca rim grew in a similar manner to that inthe andalusite and sillimanite zones, the outermostlow-Ca rim is ascribed to garnet growth during orfollowing the reaction of muscovite to form sillimaniteand K-feldspar. By analogy with the study of Pyle &Spear (2003), the Y annulus may reflect garnet growthin the presence of a melt phase, consistent with themigmatitic textures in the rocks (Pattison & Vogl,2005). In general, the Ca zoning in the sillimanite andK-feldspar zones shows more evidence for garnet dis-solution than in the staurolite and andalusite zones,most likely due to higher absolute temperatures.

INTERPRETATION OF DISCREPANCIESBETWEEN EQUILIBRIUM MODELLING ANDOBSERVATIONS

The major discrepancies between the observations andphase equilibrium predictions include the:1 relative unreactivity of garnet going upgrade;2 close proximity of the garnet-in, staurolite-in andandalusite-in isograds;3 patchy progress of the staurolite-consuming, anda-lusite-producing reaction; and4 relatively abrupt disappearance of staurolite oncesillimanite develops.

Unreactivity of garnet

The observations above suggest that garnet grew inintervals in which it was a reaction product but did



not dissolve as much as predicted when it was areactant. We therefore tested a scenario in whichgarnet is modelled to undergo perfect progradefractional crystallization during periods of predictedgrowth, but not to dissolve at all during periods ofpredicted dissolution. The fractional crystallizationcalculations utilized the program THERMOCALCTHERMOCALC andwere calculated using the predicted garnet mode andcomposition to modify the whole-rock compositionin one-degree increments to account for growinggarnet.

The calculated modal behaviour of garnet in thetwo models is shown in Fig. 17. In the fractionalgrowth model, the bulk rock becomes impoverishedin elements that fractionate into garnet, in particularMn, driving the system closer to the Mn-freeNCKFMASHT system as garnet grows. The effect isto displace the garnet-in line continuously up-tem-perature and up-pressure as garnet grows, and con-comitantly, but to a much lesser degree, the positionof the isopleths. The displaced garnet-in line for thefractionated bulk composition at the point thatstaurolite begins to grow is shown as dashed lines inFig. 12a. An interesting feature in this model is thatgarnet is not predicted to grow at all upgrade of thegarnet zone (the garnet-in isopleth lies above3.5 kbar).

Comparison of the predicted behaviour of garnetwith the observed garnet textures and chemicalzoning in Fig. 17 suggests that neither of the equi-librium nor fractional growth ⁄no dissolution end-member models satisfies the observations. There issubstantially less garnet dissolution in the staurolitezone than predicted by the equilibrium model butmore garnet growth in the andalusite zone thanpredicted by the fractional growth ⁄ no dissolutionmodel. The natural process may lie somewherebetween these end-members. On the other hand,garnet growth in the sillimanite zone is predicted byneither model, and is most likely due to the kineticdelay in the staurolite-consuming, sillimanite + garnet-producing reaction, resulting in metastable garnetgrowth at higher temperatures.

Reaction overstepping and reaction affinity

As noted above, several reactions in the aureole appearto have been overstepped in temperature. A primarydeterminant of the extent to which a reaction is over-stepped due to kinetic barriers is the reaction affinity(see Introduction). Reaction affinity in thermallyoverstepped reactions is related to the entropy differ-ence (DS) between the thermodynamically stable, butnot yet crystallized, equilibrium assemblage, and themetastable reactants, by the equation A = DT * DS,where DS is as defined above and DT is the tempera-ture overstep of the equilibrium boundary (=T ) Teq).For the prograde reactions of concern to this study,T > Teq.

Calculating reaction affinity in natural systems is aninexact exercise because the nature of the composi-tional departures of reactant and product phases fromequilibrium cannot be predicted. Some reactant phasesmay change composition as the reaction is oversteppedwhereas others may not, and the compositions ofproduct minerals that do finally nucleate and grow areunknown other than within limits.

We therefore follow the approach of Walther &Wood (1984), Lasaga (1986) and Waters & Lovegrove(2002), who calculated reaction affinities for simplifiedend-member reactions that provide an approximationto the more complex reactions that actually proceededin the rocks. Using this approach, the above equationbecomes A = DT * DrSP,T, where DrSP,T is theentropy change of the model reaction.

Table 3 lists entropy changes for the model reactionsrelevant to this study. With one exception, the entropychanges were calculated at 3.5 kbar, 560 �C, the equi-librium temperature of the staurolite-forming reaction.The entropy change of the sillimanite+K-feldsparreaction entropy was calculated at 3.5 kbar, 650 �C.The thermodynamic data of Holland & Powell (1998,2003) were used. The entropy values in Table 3 are themean of the values for the Fe- and Mg-end-memberreactions, normalized per mole of oxygen in theporphyroblast phase to allow comparison betweenreactions (Waters & Lovegrove, 2002). The normalized

Table 3. Entropies of reaction.

Reaction

number

in text

Reaction

label

Simplified reaction in KFASH or KMASH No. oxygens in

porphyroblast

Normalized

moles H2Oa

Normalized

DrSP,Tb

Width of Fe-Mg

divariant interval in �C

1 Grt-in 1Ms + 3Chl + 3Qtz = 4Grt + 1Bt + 12H2O Grt: 12 0.25 16.1 40

3 St-in 41Ms + 31Chl = 8St + 41Bt + 33Qtz + 108H2O St: 48 0.28 17.0 3

5 And-in

(from Chl)

5Ms + 3Chl = 8And + 5Bt + 1Qtz + 12H2O And: 5 0.30 19.1 2

4 And-in

(from St)

8Ms + 6St + 17Qtz = 62And + 8Bt + 12H2O And: 5 0.04 3.6 5

9 Grt-out Ms + Grt = 2And + Bt + Qtz Grt: 12 0.00 0.6c >50

7 Sil-in

(from And)

And = Sil Sil: 5 0.00 0.6 0

8 Sil+Kfs-in Ms + Qtz = Sil + Kfs + H2O Sil: 5 0.20 14.2 0

aMoles H2O in reaction, per mole of O in indicated product or reactant porphyroblast.bJ K)1 per mole of O in indicated product or reactant porphyroblast.cKFASH estimate.



number of moles of H2O produced in each reaction isalso listed, from which can be seen the strong corre-lation between entropy of reaction and moles ofH2O produced.

Reaction affinities were calculated using the equa-tion A = DT * DrSP,T. A complication in applying thevalues in Table 3 to the real reactions in the rocks isthe effect of multivariancy (for example, in KFMASH,Fe-Mg divariancy). Whereas the Sil+Kfs-producingreaction (7) is univariant, giving rise to a simple linearincrease in reaction affinity with degree of temperatureoverstep, the Fe-Mg reactions in Table 3 are divariant.In an equilibrium model, the extent of reaction goingup-temperature through a divariant interval increasesfrom 0% at the temperature at which the reactioninterval is entered to 100% at the temperature at whichthe reaction interval is exited. The entropy change atany given temperature in a divariant interval is there-fore the product of the end-member (or �full�) entropychange listed in Table 3 and the predicted extent ofreaction at that temperature. Reaction affinity, theproduct of the temperature overstep and the predicted

entropy change, therefore increases nonlinearlygoing though the reaction interval, becoming linearthereafter.Reaction affinity v. temperature for the reactions

encountered going upgrade along an isobaric P-T pathat 3.5 kbar is illustrated in Fig. 18a. The temperaturesof each reaction at zero reaction affinity (i.e. at equi-librium) are given in Fig. 12a and Table 2. Metastablereaction (5), involving growth of andalusite fromMs+Chl, is shown as a dashed line. For most of thereactions in Table 3, the width of the multivariantreaction interval is <5 �C and so is ignored for thepurposes of this study. For garnet-forming reaction(1), however, the multivariant interval is �40 �C,based on calculations in KFMASH for a bulk-rockMg ⁄ (Mg+Fe) corresponding to the average pelite.The curvature of reaction (1) in Fig. 18a going throughthe divariant interval results in significantly lowerreaction affinity at small temperature oversteps (e.g.<20 �C) compared with the situation in which themultivariancy of reaction (1) is ignored (linear dashedline in Fig. 18a). A general feature of Fig. 18a is that,

Ms + Chl to

Grt +

Bt rxn (1

)

metastable

Ms +

Chl to

And + B

t rxn (5

)

Ms + C

hl to St +

Bt rxn (3

)

Ms + St to And + Bt rxn (4)

Ms to Sil +

Kfs rxn (8

)

And to Sil rxn (7)

Reaction affinity v. temperature

Rea

ctio

n af

finity

per

mol

e of

oxy

gen

in p

rodu

ct p

orph

yrob

last

(J)

Temperature (°C)

Univaria

nt

Divaria

nt

Prograde path in the Nelson aureole

Ms + Chl to

Grt +

Bt rxn (1

)

Metastable

Ms +

Chl to

And + B

t rxn (5

)

Ms + C

hl to St +

Bt rxn (3

)

Ms+St to And + Bt rxn (4)Ms to

Sil + Kfs rx

n (8)

And to Sil rxn (7)

1500

1000

500

0

1500

1000

500

0500 525 550 575 600 625 650 675 700

Energy barrier to nucleation and growth

Small overstep of 12 °Cfor large S reaction

Large overstep of 55 °Cfor small S reaction

1. Overstepping of Grt-forming reaction: nucleation difficulty

2. Once garnet nucleates, 'cascade effect':- release of H2O into grain boundary medium- rapid growth of garnet from Ms + Chl matrix- lowered kinetic barriers to nucleation and growth of St and And from Ms + Chl matrix- abrupt loss of chlorite and release of water over small temperature interval

3. Lack of reaction of St to And, except where locally catalyzed by infiltrated fluid

4. Abrupt reaction of St to Sil:more favourable nucleation?

5. unknown (small?)overstepping of Ms-out reaction

Fig. 18. (a) Reaction affinity v. temperature for reactions in the Nelson aureole. Slopes of reactions are calculated from entropy datain Table 3, with the exception of the curved part of garnet-forming reaction (1), which is discussed in the text. The equilibriumpositions of reactions (intercepts along horizontal axis, corresponding to zero reaction affinity) come from Fig. 12a and Table 2. Thedotted line for energy barrier to nucleation and growth is representative only. See text for further discussion. (b) Representativeprograde path that accounts for many of the features observed in the Nelson aureole. The dotted arrow and curve for reaction (1) arediscussed in the text.



for a given reaction affinity needed to overcome bar-riers to nucleation and growth, reactions with highDrSP,T achieve this after a smaller temperature overstepthan reactions with low DrSP,T.

Overstepping of garnet formation

The near coincidence of the garnet-in isograd with thestaurolite-in and andalusite-in isograds, and the clus-tering of garnet core isopleth intersections around thestaurolite-in reaction in Fig. 13a are consistent withoverstepping of the garnet-forming reaction. Over-stepping is expected to result in a relatively rapidgrowth of garnet, consistent with the textural andchemical zoning features in garnet noted above.

The magnitude of the overstep in energy needed tonucleate and grow garnet can be estimated fromFig. 18b. Delay in the onset of the garnet-formingreaction to the temperature of the staurolite isograd(�30 �C) corresponds to a reaction affinity of �400 Jper mole of oxygen in garnet, equivalent to �4.8 kJ permole of garnet or �1.6 kJ per mole of H2O released inthe reaction. As noted by Waters & Lovegrove (2002),these values correspond to the lower end of the rangeof 1–10 kJ per mole of H2O deduced by Ridley &Thompson (1986) in their comparison of seeded andunseeded phase equilibrium experiments, and arecomparable with the estimates of Wilbur & Ague(2006) from crystal growth simulations. On the otherhand, if the equilibrium position of the garnet-inreaction occurs only 10–15 �C below the staurolite-inreaction, such as modelled in Figs 12 & 13, the pre-dicted reaction affinity due to thermal overstepping ismuch lower (<100 J per mole O in garnet) because ofthe shallowing slope of reaction (1) in Fig. 18b withdecreasing temperature overstep (dotted line inFig. 18b).

The most likely cause of the delay in garnet forma-tion was difficulty of garnet nucleation. Once theactivation energy barriers were finally overcome andinitial garnet growth commenced, hydrous fluidaccompanying garnet growth would have entered thegrain boundaries. This in turn would have lowered thekinetic barriers to nucleation and growth, leading to anabrupt increase in reaction progress and potentiallyrapid garnet growth.

Processes at the staurolite-in and andalusite-in isograds

In addition to the near coincidence of the garnet,staurolite and andalusite isograds, the lack of evidencefor involvement of garnet in staurolite formation andof staurolite in andalusite formation near the isogradssuggests that all three porphyroblasts grew fromreaction of matrix chlorite and muscovite over a smalltemperature interval.

There are at least two possible explanations for thisobservation. The first is a �cascade effect�, in which therelease of fluid into the intergranular domain accom-

panying the initial formation of garnet may haveabruptly lowered activation energy barriers for nucle-ation and growth of staurolite and andalusite as well.The cascade effect suggested here is somewhat differentfrom that described by Waters & Lovegrove (2002), inwhich nucleation of a phase (e.g. staurolite) up-tem-perature of where it was predicted to form allowedseveral reactions, both stable and metastable, to con-tribute to its growth, resulting in an interval of vigor-ous reaction. The fluid-related cascade effect suggestedhere may have allowed several Ms+Chl-consumingreactions to proceed in parallel. Growth of garnet,staurolite and andalusite from the same Ms+Chlsubstrate would imply that the garnet-, staurolite- andandalusite-forming reactions were all overstepped tosome degree. The relative progress of the competingreactions would then be determined by the interplaybetween the kinetics of each reaction and the overalltendency to minimize the free energy of the system.

Such overstepping opens the possibility of out-of-order mineral growth, as documented in theBushveld aureole by Waters & Lovegrove (2002) andin the Ubehebe aureole by Muller et al. (2004). Forthe Nelson aureole, we interpret the textural evidenceto indicate sequential growth of garnet, staurolite andandalusite in the order predicted in the equilibriumphase diagram. Therefore a second possible explana-tion is progressive fractional growth of each por-phyroblast mineral from the Ms+Chl matrix withoutsignificant dissolution of the earlier-formed por-phyroblasts, resulting in a progressively chlorite-poorer, Mg-richer reactant matrix composition. Thepredicted effect of this fractionation process (seefig. 12e of Pattison & Vogl, 2005) would be sequentialgrowth of garnet, staurolite and andalusite. Depend-ing on its initial abundance, loss of chlorite duringprogressive fractionation would result in the devel-opment of Grt-, Grt+St-, and Grt+St+And-bearingassemblages, as observed (Fig. 1b). Fe(±Mn)-poorerrocks may have resulted in sequential growth ofstaurolite and andalusite with no garnet, also asobserved. The lack of consumption of garnet duringstaurolite growth, and of staurolite during andalusitegrowth, may be ascribed to sluggish porphyroblastdissolution compared with the vigorous matrixreaction and porphyroblast growth in this narrowtemperature interval.

A combination of both processes may be the bestexplanation for the observations. In any scenario, thecompressed garnet, staurolite and andalusite isogradsrepresent a narrow temperature interval over whichchlorite reacted out of the rocks. The relatively abruptdisappearance of chlorite may be accounted for by acombination of a relatively steep increase in reactionaffinity as the chlorite dehydration reactions wereoverstepped, and reaction acceleration due to internalgeneration of fluid. Whereas the observed mineralassemblages match predictions from the equilibriumphase diagram quite well, the actual reactions that



produced the final mineral assemblages appear to havebeen largely dictated by reaction kinetics, althoughpossibly not very far from the equilibrium tempera-tures, with the possible exception of garnet.

The staurolite-consuming, andalusite- or sillimanite-forming reaction

In contrast, staurolite-bearing, chlorite-free rocksoccur for a significant distance upgrade of the St+Andisograd (Fig. 1b), even though staurolite is predictedto be completely consumed in the formation of anda-lusite only a short distance upgrade of the andalusiteisograd (Fig. 12a). The patchy reaction of stauroliteto andalusite in this broad interval suggests that:(i) staurolite is indeed metastable with respect toandalusite in this interval; (ii) the reaction of stauroliteto andalusite was kinetically inhibited in many rocks;and (iii) local fluid presence lowered the kinetic barri-ers to nucleation and growth, allowing the reaction toproceed locally.

These features can be explained by the low DrSP,T ofthe staurolite-consuming, andalusite-producing reac-tion (4), resulting in small increase in reaction affinitywith temperature overstep (Fig. 18). The low DrSP,T isdue to the small amount of H2O released per mole ofandalusite produced (Table 3), contrasting with thehigher vales of DrSP,T of the more substantially H2O -producing, chlorite-consuming reactions. Figure 18ashows that the equilibrium temperature of reaction (7)has to be exceeded by 50–60 �C before the reactionaffinity reaches 200 J mole)1 per oxygen in andalusite(equivalent to 1 kJ per mole of andalusite produced inthe reaction).

The reason for the inhibited reaction of stauroliteto andalusite may involve a combination of sluggishdissolution of staurolite, difficulty of nucleation ofandalusite and slow nutrient transport in rockslacking an intergranular fluid. Evidence in support ofeither or both of sluggish dissolution of stauroliteand slow nutrient transport is the occurrence ofeuhedral staurolite crystals in rocks that containandalusite (Fig. 3b), whereas evidence in support ofnucleation difficulty of andalusite is the abrupt dis-appearance of staurolite once sillimanite is involvedin the reaction. Although increasing temperature isalso expected to result in accelerated reaction ofstaurolite, the abruptness of the change suggestskinetic control related to the involvement ofsillimanite rather than andalusite in the reaction.Sillimanite is finer grained and more widely distrib-uted throughout the matrix of the rock comparedwith discrete porphyroblasts of andalusite, indicativeof easier nucleation. More widely distributedsillimanite would also favour growth because ofsmaller transport distances from reactant phase(staurolite) to product phase (sillimanite).

The importance of fluid in substantially reducingactivation energy barriers to nucleation and growth of

andalusite is apparent from the localized reaction ofstaurolite to andalusite in the wide zone between theandalusite and sillimanite isograds (Fig. 1b). Thisimplies that the grain boundaries of most staurolite-bearing, chlorite-free rocks upgrade of the St+Andisograd were volatile-deficient, indicating that theabundant H2O generated earlier in the production ofthe staurolite from chlorite consumption escaped orwas otherwise consumed. The source and drivingmechanism for the localized influx of fluid that facili-tated reaction of staurolite to andalusite is uncertain,but could be related to fluids released from thedehydrating intrusion, or channelled metamorphicdehydration fluids from below.A final feature of this domain is the evidence for

garnet growth in both the andalusite and sillimanitezones. The fact that garnet grew from a metastablestaurolite-consuming reaction in a domain where itshould instead have been consumed underscores theimportance of kinetically controlled reactions in pro-grade metamorphism, especially in situations wherethe DrSP,T of the stable reactions is small. Table 3shows that the DrSP,T of reaction (9), by which garnetis predicted to be consumed in the andalusite zone, isindeed extremely low.

Synthesis

Figure 18b provides a synthesis of the above inter-pretations in the form of a prograde path throughreaction affinity–temperature space that qualitativelysatisfies the range of spatial, textural and composi-tional features of the mineral assemblages in theaureole. The characteristic feature of the path inFig. 18b is a sawtooth pattern that illustrates theinterplay between equilibrium and kinetics in the pro-gress of the reactions: intervals of reaction overstep-ping are marked by rises along the reactions, whereasintervals of reaction are marked by steep drops downtowards equilibrium. The vertical drops to equilibriumshown in Fig. 18b imply that nucleation is the solerate-limiting factor in reaction progress, whereas ifporphyroblast dissolution and nutrient transport werealso factors, as is likely the case, the drops would beless steep.

IMPLICATIONS FOR PROGRADEMETAMORPHISM

Discrete intervals of reaction and of fluid generation inprograde metamorphism

Slowly increasing reaction affinity through broadmultivariant domains in P–T phase diagram sections(pseudosections) may mean that reaction progressgoing through these domains is less than thatpredicted thermodynamically, and is unlikely to becontinuous. Reaction may take place instead inabrupt, discrete intervals within these multivariant



domains, corresponding to the point at whichsufficient reaction affinity has built up to overcomekinetic barriers to nucleation and growth. Reactionmay not occur at all if the free energy change goingthrough these domains is small, and metastablereaction is possible. These considerations contrastwith the idea that most of the reaction that ametamorphic rock experiences along a P–T–timepath occurs in broad multivariant domains, and callinto question the notion of continuous equilibrationduring prograde metamorphism.

In the Nelson aureole, reaction was largely confinedto three relatively narrow intervals corresponding tochlorite consumption, metastable reaction of stauroliteto sillimanite, and muscovite consumption, with littleor possibly no reaction between except where locallyfluxed by fluids. Metamorphism was episodic ratherthan continuous, and fluid release (and possibly fluidpresence in general) was correspondingly episodicrather than continuous (cf. Thompson, 1983; Rubie,1986).

Nucleation difficulty and reaction overstepping

Several authors, such as Rubie (1998) and Waters &Lovegrove (2002), have argued that nucleationdifficulty is an important, possibly leading, cause ofreaction overstepping. Several aspects of this studysupport this inference, recognizing that the process ofnucleation is linked to processes of reactant phasedissolution and intergranular transport needed toprovide the nutrients to allow nuclei to initiate andgrow. Overstepping of the garnet-in reaction ismost plausibly explained by nucleation difficulty.Overstepping of the staurolite-out, andalusite ⁄ sillima-nite-in reaction, the latter only proceeding significantlyonce sillimanite was involved, may be related tonucleation difficulty of andalusite compared withsillimanite, although sluggish dissolution of staurolitealso appears to have played a role.

The estimated magnitude of nucleation-relatedoverstepping of porphyroblast growth from aMs+Chl protolith in this study (0–30 �C) is com-parable with the lower end of the 30–40 �C estimateof Waters & Lovegrove (2002), but rather less thanthe 80 �C suggested by Zeh & Holness (2003) in theirstudy of garnet growth from a Ms+Chl protolith inthe Ilesha Schist belt of Nigeria. The success of themetamorphic facies principle, and the consistency ofmetapelitic mineral assemblage sequences in whichthe temperature difference between mineral assem-blages is typically less than about 40 �C, arguesagainst such extreme overstepping of devolatilizationreactions of moderate to high DrSP,T, other than inunusual circumstances. On the other hand, for reac-tions of low DrSP,T in dry rocks, more substantialoverstepping is anticipated, as seen with the stauro-lite-out, andalusite ⁄ sillimanite-in reaction in thisstudy.

Implications for thermobarometry and P–T pathdeterminations

The results of this study indicate that the broad fea-tures of a metamorphic sequence may be accounted forby an equilibrium model, but that the details of isogradpatterns, mineral textures, and mineral compositionand chemical zoning require consideration of bothequilibrium and kinetics. For example, a hithertopuzzling aspect of Area C of the Nelson aureole(Pattison & Vogl, 2005), which occurs immediately tothe south of the study area, was the absence of anandalusite zone upgrade of staurolite development.Whereas this could be interpreted as an anomalouslylow thermal gradient adjacent to the igneous contact inthis area, a more likely explanation is that, in thesamples collected, no fluid infiltration had occurred tocatalyse the reaction of staurolite to andalusite.

Although mineral assemblages may approach whatis predicted from equilibrium thermodynamics, theresults of this study suggest that the actual reactionsleading to a given mineral assemblage may be quitedifferent from those predicted by the phase equilibria.Some may be metastable, such as the staurolite-consuming, sillimanite-forming reaction. These consi-derations highlight the potential for misinterpretationof P–T paths when textures suggesting a particularsequence of mineral growth and consumption areinterpreted solely with respect to the reactions in anequilibrium phase diagram, echoing points raised byVernon & Powell (1976) and Vernon et al. (2008).

These results also introduce a note of caution toPowell & Holland’s (2008) recommended approach toP–T estimation of metamorphic rocks, involving com-bined use of mineral assemblage stability domains inequilibrium phase diagram sections (pseudosections),measured mineral modes and compositions, and clas-sical thermobarometry. The rationale for this approachis that all should agree, within uncertainty. In theNelson aureole, by contrast, one predicts that theyshould not agree because some of the minerals (e.g.garnet and staurolite) occur outside their stability fieldand, in the case of garnet rims in the andalusite andsillimanite zones, grew metastably. Whether such ascenario could result in P–T discrepancies that falloutside the limits of uncertainty remains to be seen.

Application to regional metamorphism

Other studies of contact aureoles in which disequilib-rium processes have been suggested include Hollister(1969a,b), Manning et al. (1993), Waters & Lovegrove(2002), Muller et al. (2004) and Buick et al. (2004). Animportant question is whether disequilibrium processesare restricted to contact metamorphic settings, orwhether they extend to regional metamorphism. Thetimescales of regional metamorphism, and thereforeheating rates, are generally considered to be slowerthan for contact metamorphism, although Waters &



Lovegrove (2002) argued that the timescale of meta-morphism in the Bushveld aureole approached that ofregional metamorphism.

Waters & Lovegrove (2002) noted that, according toclassical, steady-state nucleation rate laws, the activa-tion energy for nucleation depends inversely on thesquare of the temperature overstep, in contrast toother thermally activated processes (e.g. diffusion)whose activation energies are functions of absolutetemperature but not of temperature overstep. Theresult is a sharp increase in nucleation rate over a verysmall temperature interval (a few degrees), and there-fore a small dependence on heating rate. Ridley &Thompson (1986, p. 159) made the same point inarguing for the relevance of experimentally determinedreaction oversteps to real metamorphic rocks. Anexception could be the innermost zones of contactaureoles adjacent to hot intrusions, as argued by Buicket al. (2004).

A more significant factor may be the presence orabsence of rock deformation occurring duringmetamorphism (Waters & Lovegrove, 2002).Deformation assists nucleation by building up strainenergy in the reactant phases, contributing to theenergy needed to overcome activation energy barriers,and by providing energetically favourable sites fornucleation. Deformation accompanying metamor-phism is generally thought to be more common inregional metamorphism compared with contact meta-morphism, although it is important to note that manycontact aureoles show textures and structuresindicative of deformation prior to and duringmetamorphism.

Evidence in support of lower degrees of overstep-ping in regional metamorphism compared to thisstudy comes from the common occurrence of garnetzones of significant width downgrade of stauroliteisograds in regional facies series 2b (staurolite-anda-lusite) sequences of the same pressure as this study (seeexamples in Pattison&Tracy, 1991). On the other hand,the fairly common occurrence of simultaneous �stauro-lite + andalusite� and �staurolite + kyanite� isogradsin regional metamorphic settings, such as in Augusta,Maine (Osberg, 1968; Ferry, 1980) and Mica Creek,British Columbia (Ghent, 1975), respectively, andthe common lack of evidence of instability of staurolitein kyanite-bearing rocks, may be suggestive of kine-tically controlled processes similar to those thatoccurred in the Nelson aureole. Carlson (2002)demonstrated chemical disequilibrium during garnetgrowth from a number of regional metamorphicsettings, further suggesting that disequilibrium duringregional as well as contact metamorphism deservesconsideration.

ACKNOWLEDGEMENTS

We thank J. Vogl (Univ. Florida) for his geologicalmapping and sampling of the Nelson batholith and

aureole, which led to the identification of the studyarea. R. Marr (Univ. Calgary) is thanked for hisassistance with the electron microprobe. M. Jolivet,S. Lallemand and their colleagues at GeosciencesMontpellier are thanked for their hospitality duringPattison�s half-sabbatical in 2006, when much ofthis work was done. D. Waters amiably clarified anddebated several aspects of this study. D. Waters,W. Carlson, M. Brown and especially J. Ferry providedinsightful reviews and editorial comments thatimproved the paper. This work was supported byNSERC grants 037233 to Pattison and 327218 toTinkham.

REFERENCES

Andersen, T. B., 1984. Inclusion patterns in zoned garnets fromMageroy, north Norway. Mineralogical Magazine, 48, 21–26.

Archibald, D. A., Glover, J. K., Price, R. A., Farrar, E. &Carmichael, D. M., 1983. Geochronology and tectonicimplications of magnetism and metamorphism, SouthernKootenay Arc and neighbouring regions, southeastern BritishColumbia. Part I. Jurassic to Cretaceous. Canadian Journal ofEarth Sciences, 20, 1891–1913.

Buick, I. S., Stevens, G. & Gibson, R. L., 2004. The role of waterretention in the anatexis of metapelites in the BushveldComplex aureole, South Africa: an experimental study. Jour-nal of Petrology, 45, 1777–1797.

Burton, K. W., 1986. Garnet-quartz intergrowths in graphiticpelites: the role of the fluid phase.Mineralogical Magazine, 50,611–620.

Cairnes, C. E., 1934. Slocan Mining Camp, British Columbia.Geological Survey of Canada Memoir, 173.

Carlson, W. D., 2002. Scales of disequilibrium and rates ofequilibration during metamorphism. American Mineralogist,87, 185–204.

Chernoff, C. B. & Carlson, W. D., 1997. Disequilibrium for Caduring growth of pelitic garnet. Journal of MetamorphicGeology, 15, 421–438.

Connolly, J. A. D. & Cesare, B., 1993. C-O-H-S fluid composi-tions and oxygen fugacity in graphitic metapelites. Journal ofMetamorphic Geology, 11, 379–388.

Cook, F. A., Green, A. G., Simony, P. S. et al., 1988. Lithop-robe seismic reflection structure of the southeastern CanadianCordillera: initial results. Tectonics, 7, 157–180.

De Capitani, C., 1994. Gleichgewichts-Phasendiagramme:Theorie und Software. Ber. Deutsch. Mineral. Gesellsch., 72,48. (abstract).

De Capitani, C. & Brown, T. H., 1987. The computation ofchemical equilibria in complex systems containing non-idealsolutions. Geochimica et Cosmochimica Acta, 51, 2639–2652.

Ferry, J. M., 1980. A comparative study of geothermometers andgeobarometers in pelitic schists from south-central Maine.American Mineralogist, 65, 720–732.

Ghent, E. D., 1975. Temperature, pressure and mixed volatileequilibria attending metamorphism of staurolite-kyanite-bearing assemblages, Esplanade Range, British Columbia.Geological Society of America Bulletin, 86, 1654–1660.

Ghosh, D. K., 1995. U-Pb geochronology of Jurassic to earlyTertiary granitic intrusives from the Nelson-Castlegar area,southeastern British Columbia, Canada. Canadian Journal ofEarth Sciences, 32, 1668–1680.

Holdaway, M. J., 1971. Stability of andalusite and aluminum sil-icate phase diagram. American Journal of Science, 271, 97–131.

Holland, T. J. B. & Powell, R., 1998. An internally consistentthermodynamic data set for phases of petrological interest.Journal of Metamorphic Geology, 16, 309–344.



Holland, T. J. B. & Powell, R., 2003. Activity-compositionrelations for phases in petrological calculations: an asymmetricmulticomponent formulation. Contributions to Minerology andPetrology, 145, 492–501.

Hollister, L. S., 1969a. Contact metamorphism in the KwoiekArea of B.C.: an end member of the metamorphic process.Geological Society of America Bulletin, 80, 2465–2494.

Hollister, L. S., 1969b. Metastable paragenetic sequence ofandalusite, kyanite, and sillimanite, Kwoiek Area, B.C.American Journal of Science, 267, 352–370.

Jamtveit, B. & Andersen, T. B., 1992. Morphological instabilitiesduring rapid growth of metamorphic garnets. Physics andChemistry of Minerals, 19, 176–184.

Kretz, R., 1983. Symbols for rock-forming minerals. AmericanMineralogist, 68, 277–279.

Lasaga, A. C., 1986. Metamorphic reaction rate laws and devel-opment of isograds. Mineralogical Magazine, 50, 359–373.

Little, H. W. 1960. Nelson map area, west half, British Colum-bia. Geological Survey of Canada, Memoir, 308.

Manning, C. E., Ingebrisen, S. E. & Bird, D. K., 1993. Missingmineral zones in contact metamorphosed basalts. AmericanJournal of Science, 293, 894–938.

Meth, C. E. & Carlson, W. D., 2005. Diffusion-controlled syn-kinematic growth of garnet from a heterogeneous precursor atPasso Del Sole, Switzerland. Canadian Mineralogist, 43, 157–182.

Muller, T., Baumgartner, L. P., Foster, C. T. & Vennemann, T.W., 2004. Metastable prograde mineral reactions in contactaureoles. Geology, 32, 821–824.

Osberg, P. H., 1968. Stratigraphy, structural geology, andmetamorphism of the Waterville-Vassalboro area, Maine.Maine Geological Survey Bulletin, 20.

Parrish, R. R., 1992. U-Pb ages of Jurassic-Eocene plutonicrocks in the vicinity of the Valhalla Complex, southeasternBritish Columbia. In: Radiogenic Age and Isotopic Studies:Report 5. Geological Survey of Canada Paper, 91-2, 115–134.

Pattison, D. R. M., 1992. Stability of andalusite and sillimaniteand the Al2SiO5 triple point: constraints from the Ballachulishaureole, Scotland. Journal of Geology, 100, 423–446.

Pattison, D. R. M., 2006. The fate of graphite in progrademetamorphism of pelites: an example from the Ballachulishaureole, Scotland. Lithos, 88, 85–99.

Pattison, D. R. M. & Tracy, R. J., 1991. Phase equilibria andthermobarometry of metapelites. In: Contact Metamorphism(ed. Kerrick, D. M.), Mineralogical Society of America Re-views in Mineralogy, 26, 105–206.

Pattison, D. R. M. & Vogl, J. J., 2005. Contrasting sequences ofmetapelitic mineral-assemblages in the aureole of the tiltedNelson Batholith, British Columbia: implications for phaseequilibria and pressure determination in andalusite-sillimanitetype settings. Canadian Mineralogist, 43, 51–88.

Peacock, S. M., 1990. Numerical simulation of contact and re-gional metamorphism on the Macintosh microcomputer.Journal of Geological Education, 38, 132–137.

Powell, R., 1998. Calculating phase diagrams involving solidsolutions via non-linear equations, with examples usingTHERMOCALC. Journal of Metamorphic Geology, 16,577–588.

Powell, R. & Holland, T. J. B., 2008. On thermobarometry.Journal of Metamorphic Geology, 26, 155–179.

Pyle, J. M. & Spear, F. S., 2003. Four generations of accessory-phase growth in low-pressure migmatites from SW NewHampshire. American Mineralogist, 88, pp 338–351.

Rice, A. H. N. & Mitchell, J. I., 1991. Porphyroblast texturalsector zoning and matrix displacement. Mineralogical Maga-zine, 55, 379–396.

Ridley, J. & Thompson, A. B., 1986. The role of mineral kineticsin the development of metamorphic microtextures. In: Fluid-Rock Interactions During Metamorphism (eds Walther, J. V. &Wood, B. J.), pp. 154–193. Advances in Physical Geochemistry,5, Springer-Verlag, New York.

Rubie, D. C., 1986. The catalysis of mineral reactions by waterand restrictions on the presence of aqueous fluid duringmetamorphism. Mineralogical Magazine, 50, 399–415.

Rubie, D. C., 1998. Disequilibrium during metamorphism: therole of nucleation kinetics. In: What Drives Metamorphism andMetamorphic Reactions? Special Publication 138 (eds Treloar,P. J. & O�Brien, P. J.), pp. 199–214. Geological Society ofLondon, London.

Sevigny, J. H. & Parrish, R. R., 1993. Age and origin of LateJurassic and Paleocene granitoids, Nelson Batholith, southernBritish Columbia. Canadian Journal of Earth Sciences, 30,2305–2314.

Thompson, J. B., 1957. The graphical analysis of mineral assem-blages in pelitic schists. American Mineralogist, 42, 842–858.

Thompson, A. B., 1983. Fluid-absent metamorphism. Journal ofthe Geological Society of London, 140, 533–547.

Tinkham, D. K. & Ghent, E. D., 2005. Estimating P-T condi-tions of garnet growth with isochemical phase diagram sec-tions and the problem of effective bulk-composition. CanadianMineralogist, 43, 35–50.

Tomkins, H. S. & Pattison, D. R. M., 2007. Accessory phasepetrogenesis in relation to major phase assemblages in pelitesfrom the Nelson contact aureole, southern British Columbia.Journal of Metamorphic Geology, 25, 401–421.

Vernon, R. H. & Powell, C. M., 1976. Porphyroblastesis anddisplacement – some new textural criteria from pelitic hornfels– reply. Mineralogical Magazine, 40, 787–788.

Vernon, R. H., White, R. W. & Clarke, G. L., 2008. Falsemetamorphic events inferred from misinterpretation ofmicrostructural evidence and P-T data. Journal of Metamor-phic Geology, 26, 437–449.

Vogl, J. J. & Simony, P. S., 1992. The southern tail of the NelsonBatholith, southeast British Columbia: structure and emplace-ment. Geological Survey of Canada Paper, 92-1A, 71–76.

Walther, J. V. & Wood, B. J., 1984. Rate and mechanism inprograde metamorphism. Contributions to Mineralogy andPetrology, 88, 246–259.

Waters, D. J. & Lovegrove, D. P., 2002. Assessing the extent ofdisequilibrium and overstepping of prograde metamorphicreactions in metapelites from the Bushveld Complex aureole,South Africa. Journal of Metamorphic Geology, 20, 135–149.

Wilbur, D. E. & Ague, J. J., 2006. Chemical disequilibriumduring garnet growth: Monte Carlo simulations of naturalcrystal morphologies. Geology, 34, 689–692.

Yardley, B. W. D., Condliffe, E., Lloyd, G. E. & Harris, D. H.M., 1996. Polyphase garnets from western Ireland; two-phaseintergrowths in the grossular-almandine series. EuropeanJournal of Mineralogy, 8, 383–392.

Zeh, A. & Holness, M. B., 2003. The effect of reaction overstepon garnet microtextures in metapelitic rocks of the IleshaSchist Belt, SW Nigeria. Journal of Petrology, 44, 967–994.

Received 17 October 2008; revision accepted 11 March 2009.



interplay between equilibrium and kinetics in prograde

Documents